Simultaneous Cloud Detection and Removal from Bitemporal Remote Sensing Images Using Cascade Convolutional Neural Networks

In this work, cloud detection is performed using U Net architecture and determination of cloud coverage is performed using CNN. We need to consider certain elements of the climate system to forecast the climate, one of it is the role of clouds in evaluating the climate's sensitivity to change. Here, we will determine the area covered by cloud and the weather condition at specific time. 

Before performing this, we will detect the clouded part from satellite image using pre-trained U-net Layers. Later cloud coverage area and weather will be performed using CNN techniques. Experiments showed that our proposed framework can simultaneously detect and shows the coverage area of clouds along with weather condition. The dataset is collected from http://gpcv.whu.edu.cn/data/

Keywords: Detection, Convolutional Neural Network, U Net Architecture.